log rank test of 1 variable with multiple groups

[manjusha hurry]

what would be the code to carry out a log rank test of 1 variable with multiple groups, e.g. i want to compare the survival of caucasians vs. african-americans or caucasians vs. asians all captured within 1 variable called ethnicity. thanks

[Clyde Schechter]

-help sts test-

More generally, if you are going to work with survival data, it is best that you first acquaint yourself with Stata's approach to it. Run -help st- and click on the link near the top of the page to the -st- chapter of the [ST] volume of the PDF documentation that comes installed with Stata. Read that chapter. It will familiarize you with the overall scheme that Stata uses as well as several commands that will prove helpful.

[manjusha hurry]

.

[manjusha hurry]

certainly had a check through the help file, by creating other variables and hence using the strata function. except i think i am just going in circles. i am trying to run the log rank test in var1, if var1=1 vs. var1=4 or var1=1 vs. var1=2. apology if this is a basic question. just can't figure it out.

[Joseph Coveney]

label define Race 1 "african-americans" 2 asians 3 caucasians
encode ethnicity, generate(race) label(Race) noextend
tabulate race, generate(race)

stset time, failure(fail=1)
sts test race1 if inlist(race, 1, 2)
display in smcl as text "Pr>chi2 = " as result %06.4f min(chi2tail(1, r(chi2)) * 3, 1)
sts test race2 if inlist(race, 2, 3)
display in smcl as text "Pr>chi2 = " as result %06.4f min(chi2tail(1, r(chi2)) * 3, 1)
sts test race3 if inlist(race, 1, 3)
display in smcl as text "Pr>chi2 = " as result %06.4f min(chi2tail(1, r(chi2)) * 3, 1)

[manjusha hurry]

Thank you Joseph. I'll have a go at this!

[Carlo Lazzaro]

Manjusha:
probably the first line of Joseph's helpful code should be re-written as:

label define Race 1 "african-americans" 2 "asians" 3 "caucasians"

[manjusha hurry]

Thanks for the follow-up Carlo, certainly helped. worked out well